@ai-sdk/harness

The core content of @ai-sdk/[email protected] is a dependency update, synchronously upgrading the underlying dependency @ai-sdk/[email protected].

@ai-sdk/harness and @ai-sdk/workflow-harness are responsible for providing standardized runtime and validation environments for AI workflows.

The core change in this update is the sandbox-vercel package synchronously upgrading its dependency @ai-sdk/harness to version 1.0.119.

@ai-sdk/workflow-harness maintains a version synchronization strategy with @ai-sdk/harness.

@ai-sdk/harness is the underlying component responsible for test infrastructure in the Vercel AI SDK, providing test harness capabilities for sandbox environments.

@ai-sdk/workflow-harness released version 1.0.101, which is a patch-level change (Patch Changes).

@ai-sdk/harness leans toward a testing and evaluation framework, used to verify the performance of AI models and toolchains across various scenarios.

@ai-sdk/harness is lower-level general-purpose testing infrastructure; the two are released on the same version-number cadence and share the same build and release pipeline.

In SDK projects with a monorepo architecture, when the underlying harness module changes, the upper-layer sandbox integration package needs to synchronize dependency versions to ensure consistency of runtime behavior.

If a project uses both @ai-sdk/sandbox-vercel and @ai-sdk/harness, it is recommended to upgrade both to 1.0.151 to avoid sandbox execution anomalies caused by version mismatch.

If a project has high stability requirements and tends to lock versions, it should compare the same-version notes of @ai-sdk/harness before upgrading, to confirm there are no underlying behavioral changes

AI SDK 及其周边包（如 @ai-sdk/harness）一直保持着较高频率的小版本滚动更新。

@ai-sdk/workflow-harness and @ai-sdk/harness are managed together under a monorepo architecture.

In large monorepo projects, when the version of the harness package changes, all sub-packages that depend on it will also undergo synchronized patch upgrades to ensure version consistency in the dependency tree.

@ai-sdk/harness provides abstract interfaces for test fixtures and capability encapsulation as a foundational base package, and @ai-sdk/workflow-harness builds on top of it while maintaining version synchronization.

@ai-sdk/harness-pi is a sub-package under the Vercel AI SDK project that is tightly coupled with the core @ai-sdk/harness.

The harness layer of the Vercel AI SDK adopts the adapter pattern, where the core @ai-sdk/harness package defines the standard interface contract and each specific harness implementation is responsible for mapping a particular provider's API behavior to the unified interface.

Compatibility with existing code should be confirmed before upgrading, because this update also upgrades the core package @ai-sdk/harness.

workflow-harness maintains exactly the same version number as its underlying dependency harness (both at 1.0.121), reflecting the Vercel AI SDK's adoption of a unified version management strategy.

@ai-sdk/harness provides the underlying mechanisms needed to run and test AI workflows, such as simulating model responses, intercepting streaming output, and asserting tool call results.